TECH COLUMN

7-6. Thermal Management for 3D Integration (Part 2) ― The More Semiconductors Are Stacked, the Harder the “Path for Getting Heat Out” Becomes ―

Heat

release:
7-6. Thermal Management for 3D Integration (Part 2) ― The More Semiconductors Are Stacked, the Harder the “Path for Getting Heat Out” Becomes ―

In Part 1, we looked at the structural problem that 3D integration crowds heat sources together and makes it hard for internal heat to get out, and at the approach of building heat paths by combining the Z and XY directions and using TIM, graphite, metal foils, vapor chambers, liquid cooling and so on as appropriate.

However, just thinking through such thermal design does not make it work as an actual product.
Whether the designed heat path can continue to be made in the same state on a scale of tens or hundreds of thousands of pieces.
This is the other real difficulty in thermal management for 3D integration.

How thin can TIM and insulating films be made?
Why does a 0.1 mm misalignment govern product performance itself?
If automated application and reel supply cannot be supported, volume production is not possible in the first place.
And can the performance confirmed in prototyping be reproduced in volume production?

■ In the Future, "Putting Cooling Inside the Chip" May Also Become Possible

As 3D integration becomes even denser, cooling only from the outside may no longer extract internal heat sufficiently.

One direction being researched is bringing cooling flow paths even closer to the heat source.

For example, it is the idea of providing micro flow paths in

● The package interior

● Interposers

● The vicinity of chips

and flowing coolant through them.

This is the idea that,

rather than waiting for heat to come out,

the cooling side goes to meet the heat close to the heat source.

■ The Closer Cooling Is Brought, the Harder Processing Becomes

Bringing cooling flow paths closer to the chip may make it possible to reduce thermal resistance.

At the same time, however, requirements become stricter for

● Leakage

● Insulation

● Micro flow paths

● Sealing

● Bonding

● Foreign matter

● Material compatibility

In other words,

the closer the heat source and cooling are brought, the higher the demands on manufacturing technology.

Cooling performance and manufacturing difficulty rise at the same time.

■ In 3D Integration, "Thin Components" Increase

In 3D integration, space in the height direction is extremely important.

If one component can be made 0.1 mm thinner, that much may make it possible to

● Add chips

● Add cooling components

● Make the product thinner

That is why thinner films are also demanded for

● TIM

● Insulating films

● Adhesives

● Metal foils

● Graphite

However, the thinner they become, the greater the effects of

● Wrinkles

● Stretching

● Tearing

● Burrs

● Misalignment

● Transport defects

In other words,

the evolution of 3D integration also raises the demands on thin-film processing technology.

■ The Smaller the Components, the More Important Positional Accuracy Becomes

The chip is small.

The TIM is also small.

The insulating material is also small.

The openings are also small.

If things shift by 0.2 mm in this state, the ratio of the shift to the area becomes large.

For example, the TIM sticks out from the chip.

Conversely, it cannot cover part of it.

The opening position of the insulating material shifts.

Such problems occur.

In other words, with miniaturization, the same 0.1 mm takes on a different meaning.

The smaller the product becomes, the higher the value of processing accuracy.

■ As the Number of Layers Increases, Errors Also Accumulate

For example, suppose

TIM

+

Graphite

+

Insulating film

+

Adhesive

are laminated.

If each position can shift by ±0.1 mm, errors accumulate in the final product.

Thickness also changes with

the thickness tolerance of each material

+

the adhesive layer

+

lamination pressure

In other words, it is necessary to look not only at the accuracy of a single material sheet, but at the accuracy of the finished product after lamination.

■ That Is Why High-Precision Lamination Becomes Important

When combining multiple functional materials,

they can be applied one sheet at a time in the customer’s process.

However,

applying the TIM.

Applying the insulating material.

Applying the graphite.

Applying the adhesive.

The more processes increase, the more opportunities for misalignment also increase.

That is where the idea comes in of

laminating the necessary materials with high precision in advance and supplying them.

This may lead to

● Fewer customer processes

● Improved application accuracy

● Fewer parts

● Easier automation

■ However, It Is Not That "Laminating Everything Is Good"

Laminating reduces processes.

But interfaces increase.

Adhesive layers also increase.

Thickness also increases.

Warping also becomes more likely.

In other words, it is not the case that integration always improves thermal performance.

What is needed is process design that decides what to integrate and what to keep as separate components.

■ If It Cannot Be Automated, It Cannot Be Mass-Produced

When AI semiconductors and advanced packages are mass-produced, thermal management components must also be supplied in large quantities.

People applying pieces one at a time with tweezers.

Aligning positions visually.

Peeling release paper by hand.

There are limits to volume production capability this way.

That is why component shapes compatible with

● Reel supply

● Automated peeling

● Automated application

● Image-based position correction

● Robot transport

are needed.

In other words, components with good thermal performance alone are not enough.

They must be thermal management components that can be incorporated stably and automatically.

■ Component Design and Equipment Design Can No Longer Be Separated

For example, when applying TIM automatically,

not only the material itself, but also

● Carrier width

● Pitch

● Peeling force

● Liner

● Component orientation

● Position recognition marks

become important.

In other words, the design of the TIM

also affects the design of the automated application equipment.

The more advanced 3D integration becomes, the harder it becomes to think about

materials,

processing,

equipment,

and inspection separately.

■ Inspection Also Becomes Difficult

In 3D integration, the inside cannot be seen.

After lamination, it becomes difficult to check

whether the TIM is making correct contact,

whether there are air bubbles,

whether positions have shifted,

whether there is foreign matter inside.

That is why not only

inspecting afterwards

but

processes that do not create defects

become important.

Processing accuracy.

Material management.

Application accuracy.

Process control.

Through these, the probability of internal defects occurring must be lowered.

■ Even If It Cools in Prototyping, It Will Not Necessarily Cool in Volume Production

In prototyping, a skilled person applies the TIM carefully.

Removes air bubbles.

Aligns the position.

Adjusts the pressure.

Then good thermal performance is achieved.

In volume production, however,

thousands of pieces per day.

Tens of thousands.

They are processed and applied by automated machines.

Here,

● TIM thickness

● Application position

● Pressure

● Material lot

● Temperature

● Humidity

fluctuate.

Then the thermal performance of the prototype may not be reproduced.

In other words, in thermal management for 3D integration, what matters more than peak performance

is continuing to deliver the same performance in volume production.

■ Evaluation Indicators for Thermal Design Also Change

For materials alone, evaluation is done in terms of

high thermal conductivity.

Low thermal resistance.

In actual 3D integration, however, it is necessary to look at performance

● After processing

● After lamination

● After application

● After thermal cycling

In other words, the evaluation axis must change from material performance

to performance in the mounted state.

This is an extremely important way of thinking for future thermal management.

■ The Researcher's Perspective: The Heat Problem in 3D Integration Is a Problem of "Heat Paths," Not "Materials"

Using high thermal conductivity materials makes it cool.

It is not that simple.

In 3D integration, multiple thermal resistances exist:

Chip interior

↓

Bonding layer

↓

TIM

↓

Heat-spreading layer

↓

Cooling component

Furthermore, heat flows

in the Z direction,

in the XY direction,

in multiple directions.

Therefore, rather than choosing the highest-performance material,

it is important to create the path with the lowest resistance from the heat source to the heat discharge destination.

■ The Shop-Floor Perspective: In 3D Integration, "0.1 mm" Becomes Product Performance

The material manufacturer’s catalog values are the same.

The TIM is the same.

The graphite is the same.

However, if

application position,

thickness,

air bubbles,

warping,

burrs,

pressure

differ, thermal performance changes.

In other words, in 3D integration,

material specifications and processing quality cannot be separated.

Buying high-performance materials does not solve heat problems.

High-performance materials only become performance when they continue to be mass-produced in the designed state.

■ Things to Check at the Design Stage

In thermal management for 3D integration, at least the following need to be checked.

Heat source

● Which chips generate heat?

● How much heat is generated?

● Where are the hot spots?

● Do they generate heat simultaneously?

Heat path

● Is heat released in the Z direction?

● Is it spread in the XY direction?

● Where is heat ultimately discarded?

● Are there bottlenecks in the heat path?

Interface

● Is the TIM thickness appropriate?

● Is contact pressure appropriate?

● Are there air bubbles?

● Can warping be absorbed?

● Can contact area be secured?

Insulation

● Where is insulation needed?

● Is the insulation thickness appropriate?

● Are there effects from burrs or foreign matter?

● Is thermal resistance being increased too much?

Processing

● What dimensional accuracy in mm is required?

● Is lamination accuracy sufficient?

● Can burrs be controlled?

● Can thin-film materials be transported stably?

Volume production

● Can it be applied automatically?

● Can it be supplied on reels?

● Can it be inspected?

● Can prototype performance be reproduced in volume production?

In other words, thermal management for 3D integration is not completed by thermal design alone.

Materials.

Processing.

Mounting.

Equipment.

Inspection.

Volume production.

It only works when everything is connected.

■ OTIS's Perspective: This Is One of the Areas Where "the Last 1 mm of Heat" Matters Most

Designing 3D semiconductors themselves is not OTIS’s job.

Nor is designing GPUs.

Nor is manufacturing HBM.

Nor is developing cooling equipment itself.

However, in 3D integration,

between high-performance semiconductors

and high-performance cooling equipment,

there are functional components of a few tens of μm to a few mm.

TIM.

Graphite.

Copper foil.

Aluminum foil.

Insulating films.

Adhesives.

Seal materials.

These create the last 1 mm of heat.

■ Where OTIS Should Aim

What OTIS should aim for is not the semiconductor itself.

Nor is it the cooling equipment itself.

It is the “components that connect heat” in between.

For example,

Chip

↓

Thin TIM

↓

Graphite

→→→ Enclosure

or

Chip

↓

TIM

↓

Vapor chamber

or

Chip

↓

TIM

↓

Cold plate

In between, problems of

● Contact

● Insulation

● Fixing

● Position

● Thickness

always arise.

This is an area where precision processing technology can deliver value.

■ What OTIS Can Do

At OTIS, through

● Thin TIM processing

● Heat dissipation sheet processing

● Graphite processing

● Copper foil processing

● Aluminum foil processing

● Insulating film processing

● Thermally conductive adhesive processing

● EMI material processing

● High-precision die-cutting

● Microfabrication

● High-precision lamination

● Lamination of dissimilar materials

● Partial lamination

● Half-cutting

● Reel-to-reel processing

● Supply formats for automated application

● Making composite functional components

we may be able to contribute to turning thermal management materials into parts that can be used in volume production processes for 3D integration.

■ The Roles of Material Manufacturers and OTIS Are Different

Material manufacturers develop

high thermal conductivity TIM,

high-performance graphite,

insulating films,

and thermally conductive adhesives.

OTIS takes those materials and

makes them into the necessary shapes.

Combines them into the necessary thicknesses.

Laminates them in the necessary positions.

Makes them into forms that can be supplied automatically.

In other words, it turns material performance into volume production performance.

That is where OTIS’s role lies.

■ Outside OTIS's Expertise

OTIS is not a company that specializes in

● GPU design

● HBM design

● Semiconductor process development

● TSV design

● Design of 3D packages themselves

● Cold plate flow path design

● Liquid cooling system design

● Design of vapor chambers themselves

● Dedicated CFD analysis

As needed, collaboration with

● Semiconductor manufacturers

● Material manufacturers

● Package manufacturers

● Cooling equipment manufacturers

● Equipment manufacturers

is necessary.

■ Industries Where This Technology Will Be Important

★★★★★ AI semiconductors

★★★★★ GPUs and accelerators

★★★★★ HBM and advanced memory

★★★★★ 2.5D and 3D packages

★★★★★ AI servers

★★★★☆ Communication semiconductors

★★★★☆ Automotive high-performance computing

★★★★☆ Robots and edge AI

■ Summary

The more semiconductors are stacked, the farther away the exit for heat becomes

3D integration is an extremely important technology that

brings chips closer together,

shortens wiring,

speeds up data,

and raises semiconductor performance.

However, it creates the great contradiction that

electrically, you want them closer.

Thermally, you want them apart.

The more chips are stacked,

the more heat sources crowd together.

Interfaces increase.

Internal heat has difficulty getting out.

That is why, in 3D integration, releasing heat in the Z direction alone is not enough.

It is necessary to combine multiple methods, such as

spreading heat in the XY direction.

Recovering heat near the heat source.

Bringing liquid cooling closer.

Reducing the thermal resistance of interfaces.

And what exists between all of these are thin functional components such as

TIM,

graphite,

metal foils,

insulating films,

adhesives,

and seal materials.

The more advanced semiconductors become, the less conspicuous such components become.

However,

the less conspicuous a component is, the more it governs the path of heat.

What is required in the 3D integration era is

not only “the material with the highest thermal conductivity.”

It is the technology to incorporate high-performance materials in the targeted shapes, thicknesses and positions, and to keep creating the same heat path even in volume production.

The more semiconductor performance rises,

what remains in the end is how to get the heat out.

And what makes that exit work is the last 1 mm of heat.

 

Column Supervisor: Koji Kakumoto (President, OTIS Co., Ltd.)

After studying abroad and working in planning and development at a trading company, he joined OTIS Co., Ltd. in 2011. While primarily working in the Corporate Planning Department, he has also served concurrently in manufacturing and technical divisions, and since 2018 has served as President, working to drive business growth and strengthen the organization.

 

This article is a general technical explanation intended for educational purposes and does not refer to any specific company, product, or technology.

Latest Articles

Related Articles

  • 6-8-1. Processing Design That Keeps Thermal Performance from Varying in Volume Production (Part 1) ― Cooling in Prototyping and Continuing to Cool in Volume Production Are Different Things ―

    6-8-1. Processing Design That Keeps Thermal Performance from Varying in Volume Production (Part 1) ― Cooling in Prototyping and Continuing to Cool in Volume Production Are Different Things ―

    In developing thermal management components, tempe... Read More
  • 10-10. Proposals Based on the Customer’s Process (Part 3) ― A Component Only Becomes Good When One Looks Beyond the Drawing to “How It Is Used” ―

    10-10. Proposals Based on the Customer’s Process (Part 3) ― A Component Only Becomes Good When One Looks Beyond the Drawing to “How It Is Used” ―

    In Part 1, we looked at the perspective of designi... Read More
  • 10-10. Proposals Based on the Customer’s Process (Part 2) ― A Component Only Becomes Good When One Looks Beyond the Drawing to “How It Is Used” ―

    10-10. Proposals Based on the Customer’s Process (Part 2) ― A Component Only Becomes Good When One Looks Beyond the Drawing to “How It Is Used” ―

    In Part 1, we looked at the idea of understanding ... Read More

CONTACT

Okayama

+81-867-42-3690

Tokyo

+813-6810-4830

Contact